Resumen de Dynamical mapping of protein-protein interactions
In this Ph.D. the exploration problem of the potential energy landscape in proteinprotein interactions has been studied. The large configurational and conformational space in protein associations is extremely difficult in sampling terms due to the many degrees of freedom. Models accounting for restricted flexibility were adopted, i.e. interface sidechain predictions and normal modes analysis for backbone responses to perturbations. Besides, stochastic Monte Carlo sampling has been used to explore distant equilibrium points in the energetic landscape, though avoiding the step by step exploration. Moreover, in order to allow for a more populous sampling a coarsegrained model was implemented. Though, three approaches have been combined: a rigidbody docking method (pyDock), a reduced resolution model (coarsegrained) and PELE (Protein Energy Landscape Exploration) that allowed for atomistic conformational sampling. A hierarchical docking protocol was then established to generate, refine and score proteinprotein docking candidates. Finally, by combining biological information with our sampling method in the coarsegrained space, the configurational complexity could be reduced simplifying the pose generation step.
Proteins performing electronic transferences have been an ideal application case for this approach, since the distance between donor and acceptor needs to be minimized while energetically stable poses are generated. Results have highlighted the importance of the full hydrogen bond network optimization combined with an atomistic implicit solvent model. During the scoring step of proteinprotein docking candidates the approach allowed to eliminate false positives. Additionally, the coarsegrained methodology showed good results in generating docking poses for electron transfer proteins.
